Finance is Automating the Org Chart Instead of the Cash Cycle

Finance has spent a lot of time debating whether employees will adopt AI, whether CFOs will fund it, and whether the technology is mature enough to move beyond experimentation. I am not even sure these questions are useful anymore, since we are already seeing steady AI adoption in the office of the CFO.

In Auditoria’s recent 2026 research, 66.5% of finance organizations said they were increasing their investment in AI, and the average finance team was using it across 2.43 functions, up 36% in a year. Yet only 21% said their AI initiatives had produced meaningful, measurable results.

Some theories for why this is happening could be that companies have either picked the wrong tools, moved too quickly, or failed to get enough people to use them. I don’t really buy that, however, because we have been deploying AI according to the way finance is organized, so that now accounts payable, accounts receivable, treasury, and even FP&A are smarter.

Cash does not move that way. It crosses every one of those boundaries, and I think we are starting to discover the limits of making each part of finance more intelligent without doing the same for the financial cycle they collectively manage.

We Have Made the Parts Smarter

One of the stranger gaps in modern finance is that we have built very sophisticated ways for each function to understand its part of cash. AP manages what is going out, AR focuses on what is coming in, treasury manages liquidity, and FP&A tries to anticipate where the numbers are heading. Each team can do its job well and still leave the CFO without anyone who is clearly responsible for what happens to cash as it moves between those functions.

Take days payable outstanding (DPO) and days sales outstanding (DSO). I can have an AP team improve days payable outstanding by five days and quite reasonably report that as progress, while collections slow in the same quarter and DSO deteriorates by seven days. Both teams may have done exactly what was asked of them, and both may have perfectly good explanations for their results. But the business still ends the quarter two days worse at converting working capital into cash.

Finance is very good at assigning ownership to functions, but the financial outcome is often created in the space between them.

We have built systems to automate invoices, reconcile accounts, forecast cash, manage collections, and produce reports. AI is now being inserted into those same categories, which means we risk making every part of finance more efficient without making finance itself any less fragmented.

Faster Work Is Not the Same as a Shorter Journey

One finding in our research keeps coming back to me because it captures this problem unusually well. Some 43.8% now answer requests within 24 hours, the highest level we have recorded in four years. At the same time, 44.5% of teams spend 11 or more hours every week following up with vendors, customers, and internal stakeholders, up from 29.7% a year ago. Another 13.4% spend more than 30 hours a week doing it.

Those numbers only look contradictory if we assume a faster response means the work is closer to completion. For example, a supplier answers an email, but the information remains incomplete, or a customer responds to a collections request by disputing the invoice. Finance finds the discrepancy and then must locate somebody in another part of the business who can explain it.

I think we have spent too much time measuring the speed of individual tasks and too little time measuring how long it takes to get from identifying an issue to resolving it in a way that changes the financial outcome.

AI Agents Raise the Stakes

Traditional automation has served us well at speeding up a very fragmented process. However, AI has changed the game entirely. An AI agent can start to make decisions inside the automated process. That means the objectives we give these systems matter much more than they did when software was simply moving information from one field to another.

The problem becomes more obvious once AI starts acting across finance. AP may be trying to optimize payment timing, AR to accelerate collections, and Treasury to protect liquidity, all with sensible objectives of their own. The risk appears when those systems make decisions against different definitions of success, without a shared view of the financial outcome a business is trying to achieve.

This is why the next finance AI problem is not adoption. Rather, it is deciding what the system should optimize, and what happens when locally sensible decisions conflict with the broader financial outcome. Permissions, approval thresholds, audit trails, and escalation rules are important, but they answer what an AI system is allowed to do. CFOs also need to decide what the system is trying to achieve across financial and governance rules.

Integration Is Becoming an Operating Issue

Lack of integration with existing systems is now the most cited barrier to AI adoption; our research puts it at 36.7%. Resistance to technology adoption, which previously led the list, has fallen to seventh place. Separate 2026 research from ACCA and CA ANZ found that 40% of finance professionals see difficulty integrating multiple data sources as a barrier to improving business insight.

Finance leaders are no longer spending most of their energy persuading people that AI belongs in the function. They are discovering what happens once multiple AI capabilities need to operate against the same transactions, policies, approvals, exceptions, and financial outcomes.

A useful AI capability can live inside a function. A useful finance operating model cannot.

So instead of indexing how many use cases have been deployed, start identifying how many handoffs remain between identifying a problem and resolving its financial impact. Instead of looking only at hours saved inside AP or AR, I would ask if cash is moving through the business more predictably. And instead of giving every finance function its own definition of AI success, I would ask who owns the outcome when those definitions collide.

For years, finance technology has been built around the organization’s structure, so it is hardly surprising that AI is being deployed the same way. The CFO is responsible for how those functions perform, but ultimately, the business does not judge finance by whether AP, AR, Treasury, or FP&A each hit their local efficiency targets. It cares about the financial outcome they produce together.

If AI is going to change finance in any meaningful way, that must become the unit of transformation. Otherwise, we will automate the functions, improve the local metrics, and still leave the financial outcome fragmented across the same boundaries we were supposed to be fixing.